Improving the Spatial-Temporal-Resolution of Dynamic MRI

Abstract

Magnetic resonance imaging (MRI) stands out as a unique, powerful, and valuable imagingmodality due to its combined features of non-invasiveness, absence of ionizing radiation hazards, relative safety, and exceptional capability to image soft tissues with superior contrast while revealing underlying metabolic processes. Fundamentally, MRI signals are derived from the physical phenomenon of nuclear magnetic resonance (NMR). At the microscopic level, when a nucleus is exposed to an external magnetic field, the intrinsic angular momentum it possesses interacts with that field, generating a magnetic torque that induces the precession of the nuclear spin about the direction of the field at a specific rate, known as the “Larmor frequency”. The Larmor frequency is linearly dependent on the filed strength and is unique to each type of nucleus. Macroscopically, when varying radiofrequency (RF) fields are applied on top of an existing static magnetic field across an object, the bulk spins of nuclei precessing at the Larmor frequency can be selectively “excited” from their equilibrium state. By introducing a train of spatially varying magnetic fields, the Larmor frequencies of the nuclei vary spatially, corresponding to the composite magnetic fields. This variation enables spatial encoding, a fundamental concept in MRI, wherein the spatial information of an object can be extracted by separating different frequency components of the detected signal. In conventional MRI, the proton in hydrogen serves as the primary nucleus of interest due to itsabundance in biological tissues. Imaging, therefore, relies on the ability of detecting the collective precession of protons in water, fat, and other organic molecules through the manipulation of RF pulse sequences that govern the spatiotemporal magnetic fields. The unique foundation and process of imaging formation underlying MRI have inspired extensive interdisciplinary efforts across fields such as imaging science, engineering, and medicine. Advancements span coil design, shimming optimization, RF pulse and sequence engineering, novel image reconstruction algorithms, and innovative post-processing techniques, including tissue recognition, segmentation, and classification. These developments collectively aim to enhance the diagnostic capabilities and clinical applications of MRI, maximizing its benefits for patient care. This dissertation primarily focuses on the development of a pulse sequence and a corresponding super-resolution reconstruction method aimed at improving spatial and temporal resolution while reducing artifacts in MRI. The overall structure of this thesis is organized as follows: Chapter 1 introduces the general concepts of MRI, providing foundational knowledge. Chapter 2 presents the proposed novel complex-valued spatial-temporal super-resolution reconstruction framework, detailing its development and design. Chapter 3 discusses the numerical simulations conducted to validate the efficacy of the reconstruction approach. Chapter 4 describes the implementation of the pulse sequences, and the design of the phantom used for testing. Finally, Chapter 5 evaluates the performance of both the pulse sequences and the reconstruction method through experimental studies.